Skip to content

6.2-exam3-2005-strang

Quiz 3, 2005 (strang) · Section 6.2: Diagonalizing a Matrix Source: exam3_1.pdf, pp. 94, 97

Problem

1 (37 pts.) (a) (16 points) Find the three eigenvalues and all the real eigenvectors of \(A\). It is a symmetric Markov matrix with a repeated eigenvalue.

\[ A = \begin{bmatrix} \frac{2}{4} & \frac{1}{4} & \frac{1}{4} \\ \frac{1}{4} & \frac{2}{4} & \frac{1}{4} \\ \frac{1}{4} & \frac{1}{4} & \frac{2}{4} \end{bmatrix}. \]

(b) (9 points) Find the limit of \(A^k\) as \(k \to \infty\). (You may work with \(A = S\Lambda S^{-1}\) without computing every entry.)

(c) (6 points) Choose any positive numbers \(r, s, t\) so that

\(A - rI\) is positive definite

\(A - sI\) is indefinite

\(A - tI\) is negative definite

(d) (6 points) Suppose this \(A\) equals \(B^TB\). What are the singular values of \(B\)?

Solution (a) The characteristic polynomial of the matrix \(A\) is \[ -\lambda^3 + \frac{3}{2}\lambda^2 - \frac{9}{16}\lambda + \frac{1}{16} = -(\lambda - 1) \left(\lambda - \frac{1}{4}\right)^2, \] and thus the eigenvalues of \(A\) are 1 with multiplicity one and \(\frac{1}{4}\) with multiplicity two. Clearly the eigenvectors with eigenvalue 1 are the non-zero multiples of the vector \((1, 1, 1)^T\). The remaining eigenvectors are all non-zero vectors of the orthogonal complement of \((1, 1, 1)^T\): they are the vectors \[ \begin{bmatrix} a \\ b \\ -a-b \end{bmatrix}, \quad \text{with} \quad (a, b) \neq (0, 0), \] and an orthogonal basis for this vector space is \[ \begin{bmatrix} 1 \\ -1 \\ 0 \end{bmatrix}, \quad \begin{bmatrix} 1 \\ 1 \\ -2 \end{bmatrix}. \] (b) For the matrix \(S\) we may choose the orthogonal matrix \[ S = \begin{bmatrix} \frac{1}{\sqrt{3}} & \frac{1}{\sqrt{2}} & \frac{1}{\sqrt{6}} \\ \frac{1}{\sqrt{3}} & \frac{-1}{\sqrt{2}} & \frac{1}{\sqrt{6}} \\ \frac{1}{\sqrt{3}} & 0 & \frac{-2}{\sqrt{6}} \end{bmatrix}, \] and the matrix \(\Lambda\) is then the diagonal matrix with entries \(1, \frac{1}{4}, \frac{1}{4}\) along the diagonal. We have \[ \lim_{k \to \infty} A^k = S \left( \lim_{k \to \infty} \Lambda^k \right) S^{-1} = S \begin{bmatrix} 1 & 0 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix} S^T \] and therefore \[ \lim_{k \to \infty} A^k = \frac{1}{3} \begin{bmatrix} 1 & 1 & 1 \\ 1 & 1 & 1 \\ 1 & 1 & 1 \end{bmatrix}. \] (c) Any \(r < \frac{1}{4}\) is such that \(A - rI\) is positive definite. Since we want \(r\) to be positive, we may choose \(r = \frac{1}{8}\). Any \(\frac{1}{4} < s < 1\) is such that \(A - sI\) is indefinite. We may choose \(s = \frac{1}{2}\). Any \(1 < t\) is such that \(A - tI\) is negative definite. We may choose \(t = 2\). (d) The singular values of \(B\) are 1, \(\frac{1}{2}\) and \(\frac{1}{2}\).

Keywords: Markov matrix, symmetric matrix, eigenvalues, matrix powers, positive definite